Modeling the triple-GEM detector response to background particles for the CMS Experiment

Modern particle physics collider experiments use collimated high energy particle beams to either collide them or dump on a fixed target. Due to the high collision rate and the interaction of beam particles with matter a hostile radiation environment is created. This radiation field is composed mainly of low energy neutrons, photons ($\gamma$), electrons/positrons ($e^{\pm}$) and charged hadrons namely kaon ($K^{\pm}$), pion ($\pi^{\pm}$), proton ($p^{+}$) [1]. Particles in the radiation field are commonly known as background particles and they have an energy typically below 1 GeV. Due to the large interaction cross-section those background processes can cause damage to detector elements and front-end electronics [2], additionally they can induce spurious signals that degrade detector performance. A strategy for radiation protection is crucial during the design and upgrade of LHC experimental facilities [3]. This strategy uses estimates obtained with Monte Carlo (MC) simulation and experimental measurements. Dedicated data taking campaigns are used to collect background data to understand detector behavior with respect to variations in experimental parameters such as luminosity and detector coordinates (i.e. R, z, $\phi$). Additionally new detector technologies are studied using high radiation doses in dedicated facilities, for instance the CERN High Energy Accelerator Mixed Field (CHARM) [4] and the Gamma Irradiation Facility (GIF++) [5] at CERN. A set of detectors based on triple-GEM technology [6] is being installed in the Compact Muon Solenoid (CMS) detector for Run-3 and Phase-2 (High Luminosity LHC) muon upgrade program [7] and similar technology has been adopted for the upgrade of the ALICE TPC [8]. Simulation plays a key role in understanding the impact of radiation on detector performance, predict behavior with an increased radiation dose and prepare strategies for mitigation. To study the radiation environment and its impact on detector performance dedicated simulation tools are used, those involve a description of complex physics processes and particle decay chains. The present study extends our previous work [9, 7] by discussing in depth the simulation strategy and methodology, improvement in the detector description and the estimation of detector response for a single triple-GEM detector as a function of incident particle’s kinematical properties (energy, momentum direction). This work could also be useful for other muon systems simulation studies to estimate background response [10, 11].

The present study aims to serve as a reference for detector response to background particles, including a detailed description of the simulation strategy and methodology, estimation of detector response for a single triple-GEM detector as a function of incident particle’s kinematical properties (energy, momentum direction). Variations to detector configuration are considered to show the validity of the estimations and robustness of the simulation model. Furthermore, the simulation model is compared with experimental data collected during 2018 by the GE1/1 slice test exercise at the CMS detector. Having a reliable modeling of the detector response will serve for future studies which are relevant considering the preparation for the High Luminosity LHC [12] in which the collision rate will increase one order of magnitude and the different detector technologies will experience a radiation environment never seen before.

Five superchambers were installed (discussed in detail in section 3.2) during 2017-2018, as shown in figure 4, in an exercise known as “slice test”, whose aim was to test electronic integration and experience in situ in the CMS cavern during LHC running conditions.

To recreate the experimental conditions and to build a simulation model, two components need to be considered, the first one is the simulation of the radiation environment. The radiation environment is estimated using a FLUKA [17] simulation with the experimental configuration of the CMS experiment at the LHC in which two proton beams collide with a center of mass energy of 13 TeV. The simulation models the particle interactions and transport through the different layers of the CMS detector. The second component is the detector response following a similar approach to the method described in section 2 but using the geometry of a superchamber. Both components are described in the following sections.

The hit rate is defined as the number of detectable signals generated by particles and collected by a particular detector device per unit of time and was studied for the GE 1/1 superchambers as a function of luminosity and detector area [22]. The hit rate is one of the most important variables to measure during the calibration and monitoring. In simulation to get an estimation for hit rate the sensitivity and particle flux need to be combined as illustrated in equation 4.1.

Here type is the type of particle (i.e. neutrons, $\gamma$’s, $e^{\pm}$ and charged hadrons), E is the energy of the incident particles and $\theta$ is the angle with respect to the axis perpendicular to the detector surface. The perpendicular distance from the beamline is denoted by $R$. The particle flux is obtained from FLUKA estimates presented in figure 5 top-right and multiplied by an average sensitivity. The average sensitivity is obtained by convolution of the sensitivity at a given energy and incident angle with abundance of particles at that energy and incident angle. The integration of this convolution over a given energy range gives the average sensitivity. The flux take care of all possible particles arriving from different directions at a given position at the detector while average sensitivity scales that flux into total hits at that position based on detectors response.

The simulation technique used here for sensitivity estimation is similar as described in section 2. The only main difference is that now we have a superchamber and additional material due to presence of “layer-1” in front of “layer-2” which affect layer-2’s sensitivity as its measured with respect to incident particle arriving at the superchamber surface from different directions. The sensitivity values for “layer-2” of superchamber as a function of energy of primary incident particles, arriving at superchamber surface from both sides, convoluted over all possible incident angles is shown in figure 7 for neutron, photon, electrons/positrons and charged hadrons.

The values for the average sensitivity estimated using energy spectrum of the incident background particles and sensitivity from different energies are presented in table 5 both for "layer-1" and "layer-2". The approach of using average sensitivity is preferred rather than obtaining sensitivity for different $\eta$-sectors because these detectors have uniform response for each $\eta$-sector, as verified with X-rays during quality control tests [23].

The effect of detector material and configuration on sensitivity could be realized from a comparison of figure 3 left and figure 7 as they represent basic geometry and CMS specific geometry respectively.

The accuracy on the sensitivity estimations depends on a correct description of the physics processes and on a realistic detector modeling. Physics processes considered in this simulation are well known and have been validated in Geant4 framework through the years by several studies and comparisons with experimental data [24], [25]. To quantify the impact of the detector modeling on sensitivity estimations the following parameters are varied with respect to the default configuration.

• •

  Drift Gap Width (DGW): This variable considers the possible variations in the thickness of the gas gap and GEM foils due to mechanical deformations during detector assembly. Such realistic variations [7] of $\pm$10 $\%$ are used for Drift Gap in the simulation for both layers of the superchamber simultaneously and the impact on average sensitivity of Layer-2 is shown in table 6.

• •

  Gas Mixture Proportion (GMP): This variable considers the possible variations in the proportions of Ar and CO₂ mixture during operation. Such variation were monitored during the detector operation and quality control test and found to be negligible, however a conservative variations are used in the simulation. The variation of Ar/CO₂ mixture considered here are (60/40) and (80/20). The variations are used in the simulation and the impact on the estimated average sensitivity of Layer-2 is shown in table 6.

The magnitude of these variations depends on the type of incident particle, kinetic energy and incident angle, and makes the detector less sensitive as compared to default configuration. The impact of these variations on sensitivity for different energy ranges are also estimated. This estimation is performed in three energy ranges for neutrons: low energy (LE) from $10^{-11}$ to $10^{-2}$ MeV, intermediate energy (IE) from $10^{-2}$ to $1$ MeV, and high energy (HE) from 1 to $10^{4}$ MeV. Two energy ranges for $\gamma$’s and $e^{\pm}$: LE from $10^{-2}$ to 1 MeV and HE from 1 to $10^{4}$ MeV. The maximum variations due to GMP for neutron in LE (IE/HE) range is estimated to be 1.0 (5.6/4.6) $\%$. For photon, $e^{\pm}$, and charged hardons these variations in LE (HE) range is estimated 0.8 (0.5) $\%$, 2.1 (0.3) $\%$, and 1.2 (0.1) $\%$ respectively. For DGW uncertainties similar order of variations estimated except for neutron case in IE region where low statistics dominate these variations.

The systematics arising from the method to generate a different primary source near the surface of the detector is also evaluated. The distance of source surface is varied by $2$ mm on either side of the nominal value of $3$ mm and average sensitivity is re-calculated to measure the impact on hit rate. A total variation of 0.5-0.8 $\%$, 0.0-1.0 $\%$, 0.2-0.4 $\%$, and 3.9-8.8 $\%$ on average sensitivity is estimated for neutron, $\gamma$, $e^{\pm}$, and charged hadrons respectively. These uncertainties are also summarized in table 6.

The other systematics from simulation setup comes from the variation in the $x$-$y$ dimensions of the plane used as source of incident particles. The source plane’s size, both at front and back of superchamber at a distance of 3$~{}mm$, are change by $\pm$10 $\%$ and variation in average sensitivity is measured. The maximum variation for a 10 $\%$ larger size of source plane is found to be about 1.3 $\%$ or less for any given type of incident particle. For a 10 $\%$ smaller size of source surface give a maximum variation of 6.7 $\%$ or less.

The uncertainties described in table 6 for different types of incident particles directly translate to uncertainty on the hit rate as per their relative contribution to total hit rate. The final uncertainties on hit rate is dominated by neutron and photon contributions.

The systematic uncertainty associated with the particle flux is estimated comparing two CMS-FLUKA geometry scenarios both of them consistent with the Run-2 data taking scenario, the uncertainty was evaluated comparing the particle flux between the two scenarios and the result is a variation of 10 to 20 $\%$ as a function of R with a mean value of 15 $\%$, the latest is considered as the associated uncertainty.

The final combined uncertainty from FLUKA simulation and Geant4 detector simulation, including various parameters, is estimated to be about $\sim$14.5 $\%$ on hit rate, after adding them in quadrature for the whole range of R considered in this study.

A comparison between simulation and experimental measurements can be used to validate the model presented in this study. Figure 8 compares the measured hit rate from data of layer-2 of superchamber 28 from CMS experiment with the prediction of simulation. The simulation results are obtained by taking the convolution of average sensitivity shown in table 5 and values shown in figure 5 top-right. The different data points correspond to different radial distance from the beamline i.e, $z$-axis, to the centre of the different $\eta$-sectors. As expected the hit rate tends to be higher in lower the $R$ (higher $\eta$) region as flux from collision is primarily directed towards this region. The bottom plot of figure 8 shows the ratio of data and simulation along with the estimated systematic uncertainty. The contributions from different types of particle to total hit rate is shown in figure 9 for layer-2. The largest contribution comes from neutrons while photons contributes about $\sim$ 15 $\%$. Charged hadrons and $e^{\pm}$ contributes about 1$\%$ only.

The simulation model from FLUKA+Geant4 discussed in this paper describes the data hit rate well within the total uncertainties, shown as orange and blue band around simulation prediction. The total uncertainty includes both statistical uncertainty and all the systematics described in previous section added in quadrature. The uncertainty on data for hit rate includes the statistical uncertainty and uncertainties on the luminosity measurement.

The simulation prediction for first two $\eta$ sectors show a larger discrepancy with respect to data most likely due to a higher threshold set during data taking operations, such effects are difficult to describe with current simulation. In future study the framework can be improved by adding a detailed description of the electron multiplication (avalanche) using GARFIELD software and a simulation of the electronics.

Environmental background hit rates on the CMS triple-GEM chamber in pp collisions at the LHC are evaluated by modeling radiation environment and detector response using a framework of FLUKA and Geant4 simulation packages. FLUKA simulation provides incident particle’s kinetic energy and angle in radiation environment. Geant4 simulation models particle interactions based on accurate material description of GEM chambers with an operation condition. The hit rates are obtained by combining sensitivity and particle flux, and compared with measurements at a luminosity of $1.5\times 10^{34}$ cm^-2 s^-1 at 13 TeV. The predicted hit rates and experimental data agree within its uncertainties. This modeling is generic, so it can be used for evaluation of hit rates on other detectors at higher luminosities, providing relevant information for detector design and operation at High-Luminosity LHC.